Automated electrical machine design with differential evolution techniques

Abstract

The design process of electrical machines is an iterative process consisting of a sizing phase and an optimization phase. The common use of finite element methods during the optimization phase results in a time consuming development, where only short user inputs are required. This harms the design experience especially for the developer, who spends more time waiting for simulation results than for deciding how the machine can be improved. This paper introduces an automated design process which autonomously computes the optimal machine geometry in dependency of an individual objective function. Thereby the involvement of the developer in the optimization phase is reduced, while his knowledge is included by weighting factors in the objective function.